Research
1. Bone Water as an Imaging Biomarker of Bone Fragility and Therapeutic Target
Bone fracture resistance has long been focused on its combination of compliant and stiff materials (collagen and mineral, respectively). Remarkably and historically undervalued, bone water (10-20% by volume) exists in a bound state throughout its multiscale hierarchy, supporting the collagen-mineral interface and imparting structure to the individual collagen triple helices and mineral crystals. This intricate composite of mineral, collagen, and water endows bone with the compressive strength of steel while being over three times lighter and exhibiting exceptional viscoelastic properties, enabling creep and stress relaxation. An arm of our group’s research explores bone water as a biomarker and therapeutic target. By combining advanced spectral techniques such as Fourier transform nuclear magnetic resonance spectroscopy (ssNMR) with thermal manipulation, mechanical testing, and modeling, we aim to spatially resolve bone's distinct water pools during aging and disease, providing new insights into bone health and potential therapeutic interventions.
2. Leveraging Radiomics and Machine Learning for Enhanced Bone Fracture Prediction
Our group aims to identify imaging biomarkers of risk and resilience to musculoskeletal decline in aging and fracture-prone conditions, including chronic kidney disease (CKD). We are advancing radiomics and machine learning techniques to extract critical imaging features, transforming high resolution peripheral quantitative computed tomography (HR-pQCT) and ultra-short echo time MRI (UTE MRI) images into rich, quantitative datasets. Our methods have demonstrated earlier detection of musculoskeletal changes missed by traditional imaging approaches, such as dual-energy X-ray absorptiometry (DXA), which relies solely on mass and density. For instance, radiomics analyses distinguished CKD bone from controls with AUROC values nearing 0.99. Beyond the bone, we are advancing our ability to simultaneously track muscle quality by analyzing the contribution of soft tissues around bone to improve fracture risk prediction and assess fracture healing stages. tracked muscle quality improvements following kidney transplant.
3. Integrating AI for Motion Correction and Reconstruction in Medical Imaging
Our research focuses on integrating advanced machine learning, deep learning, and artificial intelligence (AI) to improve qualitative and quantitative medical imaging. One key area of interest is resolving rigid and non-rigid motion artifacts in imaging techniques such as High-Resolution Peripheral Quantitative Computed Tomography (HR-pQCT) and Ultrashort Echo Time Magnetic Resonance Imaging (UTE-MRI), leveraging insights from established technologies like computed tomography (CT) and magnetic resonance imaging (MRI). We also explore sparse-view CT reconstruction and under-sampled MRI reconstruction to reduce acquisition times without compromising diagnostic quality, utilizing techniques like compressed sensing and AI. Additionally, we are advancing Magnetic Resonance Fingerprinting (MRF), a groundbreaking MRI technique that captures unique signal patterns—similar to fingerprints—cutting imaging time by over 75%. Our research emphasizes deep learning models, particularly generative approaches, to reconstruct medical images and validate these reconstructions from a biological perspective.
